The identity of a mysterious Go player named “Master” that has been beating some of the strategy game’s elite online has been revealed.
Early last month, Master was unmasked as an updated version of AlphaGo, an AI program designed by Google’s DeepMind unit in London.
Last March, AlphaGo made history when it beat Lee Sedol, South Korea’s top Go player, in four out of five games.
The South China Morning Post reported that Ke Jie, China’s No. 1 player, had responded by posting on social media: “AlphaGo can defeat Lee Sedol, but it can’t defeat me.”
Pride comes before the fall.
It turns out that Jie was wrong, since he was among the 60 scalps that Master claimed before outing itself as an AI.
— Demis Hassabis (@demishassabis) January 4, 2017
You may be wondering how an AI beating us mere mortals in a board game matters.
Such victories further cement the fact that AI is now way beyond us when it comes to solving complex yet defined problems. This could result in the technological singularity – the time when computers outmatch human intelligence – arriving quicker than we thought.
After losing, Jie reportedly took to social media again, posting:
I have studied Go software for over half a year since March, learning theories and putting them into practices countless times. I only wondered why computers are better… Humans have evolved in games in thousands of years—but computers now tell us humans are all wrong. I think no one is even close to know the basics of Go. (Translated from Chinese)
Go, invented around 2500 years ago in ancient China, is known for being simple yet extremely deep. In a standard game, players compete on a 19×19 grid, placing black and white stones in an effort to accumulate territory.
Along with Chess, Go is considered an interesting challenge for AI that must make millions or even billions of calculations to formulate a decent move. In fact, it was recently estimated that there are:
legal positions in 19×19 game of Go. That’s significantly greater than the number of atoms in the known universe. Whoa.
AI already eclipsed humans in Chess back in 1996-1997, when reigning world champion of the time – perhaps the greatest player ever – Garry Kasparov was defeated by IBM’s Deep Blue computer.
Computers have become even better at Chess since ’97 and today’s world champion would have absolutely no chance.
To put it in perspective, consider that Chess players are ranked according to something called an ELO rating. The higher your ELO, the better you are.
In the world of Chess, 400 ELO points is the difference between an amateur and a master. See what happens when Carlsen takes on New York Chess hustlers (yes, that’s a thing) who have an ELO between 400-1000 lower than him:
AlphaGo was bought by Google in 2014 and at first trained the same way as the early Chess computers – by studying a huge database of moves made by human experts. This database gave the AI the ability to evaluate moves by itself, and later think up moves that humans cannot see.
Unlike us humans, computers lack this thing we call pride. They don’t care when they lose, and they don’t get the concept of truly fighting for something.
It’s nice to know that – at least for now – we still have the power to switch them off if they get too cheeky.